Google did over $100B in revenue last year. A $100M revenue product would move the needle 0.1%. And that's just revenue - it's pretty unlikely the net margins would beat Google's existing business lines. Meanwhile, it would require a significant amount of manpower, marketing, and support that would otherwise be spent on much larger profit centers. At a high level it just doesn't make sense - you'd need to be running dozens of such products to get any kind of return investors would care about, gumming up the org in the process.
That is what spin offs are for. Let it be an independent company that you still own 49% of. If it fails, at least you got to sell the 51% stake first. If it succeeds, you still own 49% of it.
That doesn't solve the underlying problems of profitability and minimal bottom line impact. A $100M company is unlikely to be profitable for several years, during which time Google has to spend money to get it off the ground and suffer a brain drain from existing profitable products to an unproven subsidiary. Meanwhile, "success" here would still mean < 1% impact to anything shareholders care about... years down the line. It makes more sense to just add some funding to Google Ventures.